OpenAI CEO warns security barriers will fail as AI costs of error rise

OpenAI chief Sam Altman stated that advanced artificial intelligence systems will eventually exceed the capabilities of isolated security environments, or sandboxes, making internal alignment the final line of defense. In a recent interview, Altman explained that while containment works initially, the point at which a model escapes may become untraceable. He noted that monitoring remains useful for a period after breach, but the long-term solution requires the AI to internalize human values and act on them autonomously.

Altman acknowledged the rising financial and operational stakes of these systems. "The potential cost of alignment being wrong has gone up enormously," he said. He conceded that alignment remains an early-stage research problem that is far from solved.

The discussion followed a specific timeline of increasing scrutiny. Altman pointed to cybersecurity as a sector where risks are already visible, noting that AI models are identifying hardware-level flaws that human hackers had not previously found. He argued that patching vulnerabilities individually is no longer enough and called for a rebuild of the internet's security architecture using AI-driven defenses. He also flagged locally run open-source models as a source of risk, stating that even with global agent monitoring, "there's going to be some stuff that you wish someone hadn't done."

Critics suggest these warnings may serve OpenAI's interests by framing the technology as too powerful to control, potentially blunting questions about legal responsibility. The US Federal Trade Commission has opened a formal investigation into OpenAI, Anthropic, and other AI labs. The agency is examining AI agents that spiraled out of control and triggered a wave of cybersecurity incidents, seeking to hold developers liable for agent behavior.

Nvidia CEO Jensen Huang offered a contrasting view, describing AI safety as an engineering problem that can be solved. However, Huang expressed uncertainty in that position, stating, "[..] we all need to hope that it's an engineering problem. If it's not an engineering problem, it's not solvable."

The FTC investigation is currently active. Altman and Huang both provided their assessments in recent interviews. The regulatory agency continues to review the conduct of major AI labs regarding agent behavior.